Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Long-context evaluation on RULER 8k

91.07Score

QUOKA

9.253230.494151.73572.9759Feb 9, 2026
Updated 1mo ago

Evaluation Results

MethodLinks
2026.02
91.07
2026.02
80.19
2026.02
79.94
2026.02
79.19
2026.02
74.27
2026.02
72.46
2026.02
70.35
2026.02
68.97
2026.02
66.44
2026.02
65.36
2026.02
65.19
2026.02
64.2
2026.02
61.14
2026.02
60.88
2026.02
60.8
2026.02
60.09
2026.02
59.87
2026.02
57.62
2026.02
56.5
2026.02
54.49
2026.02
54
2026.02
51.6
2026.02
50.17
2026.02
49.23
2026.02
42.75
2026.02
35.98
2026.02
31.45
2026.02
31.1
2026.02
28.76
2026.02
27.15
2026.02
20.68
2026.02
20.21
2026.02
17.15
2026.02
13.7
2026.02
12.4